The therapy of metastatic colorectal cancer is largely unsatisfactory and new treatments or improved patient selection for the currently available treatments are greatly needed. DNA repair genes represent viable candidates as novel prognostic and predictive factors, because they are intimately connected with colorectal cancer biology and pathogenesis, and may also affect response to DNA-damaging chemotherapy. In a yeast two-hybrid screening with the mismatch repair protein MLH1 as "bait", we cloned MED1, a novel human DNA repair gene. The MED1 protein functionally interacts with MLH1 in human cells, is homologous to bacterial DNA repair glycosylases/lyases, and binds to CpG-methylated DNA, suggesting that cytosine methylation plays a novel role in mammalian DNA repair. MED1 displays thymine and uracil glycosylase activity, removing these bases from G:T and G:U mismatches. This suggests that MED1 may counteract mutagenesis by spontaneous deamination of 5-methylcytosine and cytosine to thymine and uracil, respectively. We also found that MED1 efficiently removes from DNA 5-fluorouracil, the mainstay of systemic therapy of colorectal cancer. Recently, we obtained evidence that the MED1 gene is frequently mutated (25% of the cases) in human colorectal and extracolonic carcinomas exhibiting microsatellite instability. In colorectal cancer specimens, we also detected loss of heterozygosity at the MED1 locus. The hypotheses underlying the present project are: that MED1 is an important component of the machinery ensuring genomic fidelity in human cells; that MED1 mutations are relevant to the pathogenesis of colorectal cancer; and that it may play a role in resistance/sensitivity to chemotherapeutic agents. Experiments are proposed to address: 1) the nature and frequency of MED1 mutations in colorectal cancer; 2) whether MED1 mutations are a prognostic factor and influence response to treatment.